tests for AD using dual numbers about forwarddiff.jl HOT 9 CLOSED

Sure, I can take a look at this.

By the way, is there any reason to not support Julia 0.2 right now? REQUIRE currently has julia 0.3-.

from forwarddiff.jl.

Yes, we should support Julia 0.2 as well. It is just my bad habit to work only with the latest build. Let's change REQUIRE in the next commit.

from forwarddiff.jl.

@StefanKarpinski, just to verify: does the Pkg code care if supported versions of a package are sequential?

from forwarddiff.jl.

If it does, we can just clear the old versions from METADATA.

from forwarddiff.jl.

It won't let you use it to tag things non-sequentially, but if you tag by hand, it will work. You can also clear older versions.

from forwarddiff.jl.

I renamed autodiff to something more descriptive and added a few tests for this.

from forwarddiff.jl.

That's great Miles, the dual_fad name that you chose seems better indeed. Thanks for adding the tests! Shall we close this issue?

from forwarddiff.jl.

Sure

from forwarddiff.jl.

@mlubin, following your preferred style of your code, I generalized the interface and put it in the api.jl file. Your code is nearly intact, I only had to make a couple of arguments optional so that they can interact with the generic API functions. Things start looking much more functional now... I also made tensor FAD possible, but will need some more work to overload more functions other than +,-,* (will do this).

from forwarddiff.jl.